(635d) Untargeted Metabolomics Analysis on Endoc-?h1 Cells Exposed to Bisphenols

Bisphenols, including BPA, BPF and BPS, are chemical compounds used in consumer products likes toys, food containers and electronics. They have been linked with various endocrine and metabolic diseases such as Type 2 diabetes, obesity, non-alcoholic fatty liver disease and metabolic syndrome. According to literature bisphenols can affect energy homeostasis by increasing insulin content in cells in an estrogen receptor α (ERα) dependent manner and impacting calcium signalling within α-cells, with the pancreas being a major target organ. BPA can reduce ion channel activity in β-cells, leading to increased glucose-induced calcium oscillations and insulin resistance. Even at low doses, BPA can induce greater insulin secretion than glucose. Bisphenols can also disrupt pancreatic function and glucose homeostasis by interfering with the production of adipokines, which are involved in insulin resistance. BPF has been found to decrease insulin sensitivity and impact β-cell function. Untargeted metabolomics analyses have been proven to be a promising approach for detecting biomarkers of various metabolic diseases. The pancreas is a major target organ for EDCs and is therefore frequently used as a model for testing EDCs, including bisphenols. The pancreas plays a critical role in energy metabolism and homeostasis through the secretion of hormones, such as insulin and glucagon. Here, EndoC-βh1 cells, were used to investigate the metabolic effects of BPA, BPF, and BPS, in parallel, to assess the impact that these chemicals have on insulin secretion functions in the presence of high and low glucose levels and to investigate the gaps in current knowledge about metabolic diseases.

Global untargeted metabolomics analysis on the pancreatic cells exposed to the three bisphenols was performed using an Agilent 1290 infinity UHPLC System coupled to an Agilent 6540 HRMS-QTOF in both positive and negative ionization modes. Additionally, two analytical techniques, RP and HILIC were used in order to increase the coverage of the detected metabolites. The data were acquired between 50 and 1700 m/z at a scan rate of 1.5 spectra/sec in centroid mode at a resolution of 40,000 FWHM. The source conditions were as follows: gas temperature 300^oC, drying gas 7 L/min, nebulizer 50 psig, fragmentor 250 V, skimmer 65 V, and capillary voltage 3500V or -3500V in positive and negative modes, respectively. The Agilent MassHunter Software v.B.06.01 was used to collect the data, followed by the data pre-processing. Briefly, the raw data acquired from the analyses were translated into the open format .mzML using the msConvertGUI tool included in the ProteoWizard toolkit. To initially evaluate the data, a base peak chromatogram (BPC) was plotted for each file, and boxplots representing the distribution of total ion currents per file for all samples were created. These boxplots could help identify problematic or failing MS runs. Following the initial data inspection, peak detection was performed using the centWave algorithm. This algorithm combines density-based detection of regions of interest in the m/z domain and a Continuous Wavelet Transform (CWT) based approach for chromatographic peak resolution, with optional Gauss-fitting in the chromatographic domain. After optimizing the parameters, the next step involves the chromatographic peak detection using QC samples. To obtain a global overview of peak detection, the frequency of identified peaks per file along the retention time axis was plotted helping to identify time periods with a higher number of peaks and evaluate their consistency across samples. Boxplots were also used for evaluation purposes. Finally, the optimized parameters were applied to all samples for chromatographic peak detection.

Next, the obiwarp method was employed to align the samples, using a binSize of 0.6. The settings were adjusted for each experiment, and the alignment of internal standards in QC samples was assessed. To evaluate the alignment impact, base peak chromatograms (BPC) were plotted using pre-adjusted and adjusted data, and the differences between adjusted and raw retention times per sample were plotted using the plotAdjustedRtime function. The final step was the correspondence matching, which involves the groupChromPeaks method in XCMS, and aims to group chromatographic peaks between samples into a feature. This algorithm combines peaks based on their density along the retention time axis within small slices along the mz dimension. The binSize, bw, and minFraction parameters were set to 0.25, 20, and 0, respectively.

Moving forward the detected features were reduced by removing variables with over 80% missing values based on QC samples and disregarding solvent samples with 0% missing data due to contamination. Features with RSD above 30% within QC samples were excluded. Median RSD for detected features and internal standards were calculated to evaluate process and instrument variability. Quantile normalization was used to reduce systematic and experimental variation, and a baseline to the median of QC samples was applied for normalization and Principal component analysis (PCA) was used to assess the quality of the analytical system performance and detect possible outliers. The R package xMSannotator was used for network-based annotation, which uses an integrative multi-criterion scoring algorithm for metabolite identification. Multi-level annotation was employed, comparing the MS/MS from analyzed samples with online and in-house databases. The adduct list used for database matching was "M+H", "M+H-2H2O", "M+H-H2O", "M+Na", "M+H+Na", "M+CH3OH+H", "M+K", "M+H+K", "M+H+HCOOH", "M+Na+K", "M+Na+HCOOH", "M+K+HCOOH", "M+ACN+Na", "M+ACN+H", "M+ACN+K" for positive ionisation and "M-H", "M-2H+Na", "M-H-H2O", "M-H+Cl", "M+Cl", "M-2H+K", "M+Cl+HCOOH", "M+Br", "M-H+HCOOH", "M-2H+NH4", "M-2H+K", "M+Hac-H" for negative ionisation., with mass tolerance to be set based on optimized mass errors. Candidate feature confirmation was performed by comparing fragmentation patterns of MS/MS spectra from the HMDB, KEGG, LipidMaps databases and the in-house library of analytical standards.

Since variance homogeneity and the assumption of normally distributed data were not met, a non-parametric test was selected for the data analysis. Statistical analysis was performed using Kruskal-Wallis and a Benjamin Hochberg FDR multiple testing correction in order to determine the statistically significant differential metabolites. A p-value cut-off < 0.05 was applied. Moreover, a fold-change analysis was performed, setting as a threshold at 2. Fold-change was calculated as the ratio between control samples and cells exposed to bisphenols. The global untargeted metabolomics approach helped facilitate the detection and identification of 3411 metabolic features in bisphenol-exposed pancreatic β-cells. Of these, 1479 were detected from the RP analysis, and 1932 were detected from the HILIC analysis. After statistical analysis, 29 significantly differentially detected features were identified across all chemical treatments and glucose conditions. These features were mostly metabolites from the sterol lipids, glycerophospholipids, and fatty acyl categories. Additionally, based on fold-change analysis, 814 features were annotated, of which 536 resulted from the HILIC and 278 from RP analysis. Most of the annotated metabolites from both analyses were lipids, specifically glycerophospholipids, fatty acyls, and sphingolipids.

A pathway over-representation analysis was performed using Fisher’s exact method to obtain the probability of differential metabolites being enriched with a particular pathway. The databases used were: KEGG, WikiPathways, Reactome, HumanCyc, EHMN, PharmGKB, SMPDB, BioCart, INOH, and PID. The metabolite names, KEGG IDs, and PubChem IDs and HMDB IDs were used as identifiers. Moreover, a background list, which contains all the features that have been measured by the analysis, was used. The background list is important when there is a small percentage of the detected metabolites in comparison with those in the database. The differentially expressed metabolites from RP and HILIC were involved in 87 and 49 statistically significant pathways, respectively. There were no common pathways detected among the treatments resulting from the RP analysis. However, six common pathways were identified among the treatments resulting from the HILIC analysis, which were Metabolism of lipids, Sphingolipid metabolism, Fabry disease, Globoid Cell Leukodystrophy, Metachromatic Leukodystrophy (MLD), and Oligodendrocyte specification and differentiation leading to myelin components for CNS.

Overall, the results from the low glucose experiments revealed that exposure to bisphenols significantly affected overall metabolism and the metabolism of lipids, especially sphingolipids (data not shown). The study also evidenced that BPA and BPF exposure had a higher impact on amino acid metabolism as well as glycolysis and gluconeogenesis pathways. Exposure to BPS led to the dysregulation of metabolites belonging to the family of vitamins, cofactors, and carbohydrates, leading to the perturbation of the relevant metabolic pathways. With the increase of glucose levels and the secretion of insulin by the pancreas, it was noticed that exposure to bisphenols mainly affected the metabolic pathways of amino acids metabolism, gluconeogenesis, glucose metabolism, and sphingolipid metabolism. Moreover, BPF and BPS treatments resulted in the dysregulation of metabolites belonging to the nucleic acids and peptides. To conclude, the results of the present study provide important information about the potential biomarkers and the relevant toxicity of three ubiquitous bisphenols in pancreatic β-cells. The results provided herein will be further investigated in conjunction with transcriptomics data in a multi-omics experiment to uncover hidden associations between omics variables, detect metabolic pathways affected by the exposure to these three bisphenols, and build novel Adverse Outcome Pathways (AOPs) as well as an integrated approach for testing and assessment (IATA) system to facilitate the hazard characterization of other endocrine disrupting compounds.